[darcs-users] pull performance issues with a lot of conflicts.

[Karel Gardas]

On Sun, 21 Nov 2004, David Roundy wrote:

> On Sat, Nov 20, 2004 at 05:00:59PM +0100, Karel Gardas wrote:
> > I'm replying with other status: darcs finally finished succesfully and it
> > took:
> >
> > real    348m2.581s
> > user    317m22.352s
> > sys     0m38.891s
> >
> > as expected there are conflicts in configure file presented, concretely
> > there are 451 conflicts there. My offer to provide both repositories for
> > debugging this issue is still valid. Personally I would apreciate if
> > someone with darcs source code/haskell knowledge look at this, since IMHO
> > this is not that rare scenario.
>
> Alas, this is a known issue, and is being worked on.

That's good news!

> The workaround is to avoid conflicts, which is a good idea anyways,
> particularly with darcs, where conflicts will impair your flexibility in
> reordering patches.
>
> Keeping generated files under version control is rarely, if ever, a good
> idea, and with darcs it's even worse.

Hmm, yes, storing generated configure in SCM is useless, but
unfortunatelly our project works this way and my current SCM (Arch) does
not "complain" about it with so long commit/update/star-merge operation.

FYI: I'm just currently using darcs as a "front-end" and tla as a SCM
"back-end", i.e. darcs for local hacking, pushing to repo which is
synchronized with tla archive (by hand).

Cheers,
Karel
--
Karel Gardas                  kgardas at objectsecurity.com
ObjectSecurity Ltd.           http://www.objectsecurity.com